Skip to main content
Michael Morakis

The guide

Fast website design in 2026 — the working guide.

A comprehensive guide to fast website design in 2026: what it means, why almost no one has it, what it costs you when you don't, and how to fix it. Written from the receipts of 21 rebuilds into the top 5% of the web.

This is not a listicle and it is not a pitch. It is the reference document I wish existed when I first started rebuilding slow premium-brand sites in 2019. If you read one page on this site, read this one. If you already know the material, skip to the sections that answer your specific question — the table of contents is below.

What "fast" actually means in 2026

"Fast" is not a feeling. It is three numbers that Google publishes and updates. These three numbers are called Core Web Vitals, and together with a fourth score they define the modern definition of a fast website. Everything else on this page is downstream of them.

Largest Contentful Paint (LCP)

The time from the user requesting your page to the largest visible element finishing its render. On mobile, "good" is under 2.5 seconds; "poor" is over 4 seconds. Most business websites in 2026 load their LCP at 4–7 seconds. A rebuild into the top 5% ships an LCP under 1.2 seconds. Mine consistently ship under 0.8.

Interaction to Next Paint (INP)

Replaced First Input Delay in March 2024. Measures how long the page takes to visually respond after a user taps or clicks. "Good" is under 200 milliseconds. WordPress sites with heavy plugin stacks routinely hit 400–800ms. A rebuild ships under 100ms.

Cumulative Layout Shift (CLS)

A dimensionless score for how much the layout jumps around as the page loads. "Good" is under 0.1. Ad-heavy sites and sites with late-loading web fonts routinely score above 0.3. A properly built site scores zero, because nothing shifts.

The PageSpeed Insights composite score

Google publishes a single composite score at pagespeed.web.dev. It is a weighted average of the three Core Web Vitals plus supplementary metrics (First Contentful Paint, Total Blocking Time, Speed Index). A score of 90–100 is "good"; 50–89 is "needs improvement"; 0–49 is "poor." I aim for and guarantee 95+ on mobile — mobile is the harder scoring surface, and if you win there, you have won everywhere.

Why fewer than 5% of websites qualify

According to the HTTP Archive Web Almanac (2024), less than 5% of tracked websites score 95+ on PageSpeed mobile. This number has been remarkably stable for years. The reason is not that fast websites are technically hard — the reason is that the incentives of the web development industry do not reward speed. They reward the appearance of features.

Three specific mechanisms produce this outcome:

1. Agencies bill by the feature. If your agency charges €5,000 for a booking widget, they have no incentive to tell you that the widget adds 800ms to your LCP. Their P&L wants the feature to ship. Your P&L wants the visitor to convert. Those two P&Ls diverge, and performance is what gets sacrificed.

2. WordPress is a plugin economy. The average WordPress business site in 2026 runs 20+ plugins. Each plugin adds its own JavaScript, its own CSS, its own database queries, and often its own third-party tracking script. The site owner sees "features I need." The browser sees a JavaScript bundle four times larger than the entire page's actual content.

3. Nobody checks after launch. A site launches at PSI 78. Marketing adds a heat-map script. HR adds a chat widget. Sales adds a demo booking iframe. A year later, the site is at PSI 34, and no single person changed anything they thought was significant. Death by a thousand cuts.

What slowness costs you

Slowness is not an engineering problem. It is a P&L problem, a search rank problem, and a trust problem — usually all three at once.

Conversions

Google and Deloitte jointly published research in 2020 showing that a 0.1-second improvement in mobile load time lifts retail conversions by ~8%, travel bookings by ~10%, and lead-generation form completions by ~7%. Amazon's older internal study — still the classic reference — found that every 100ms of latency cost them 1% in sales.

A brand doing 500 conversions per year at €400 average order value nets €200,000 in revenue. A 0.1-second speed lift adds ~35 conversions per year — €14,000 of new revenue for a single-time engineering intervention. Now compound that across ten of those 0.1-second improvements, which is what a rebuild delivers.

Search rank

Core Web Vitals have been a confirmed Google ranking factor since 2021. Google is explicit: "When many pages match a search, page experience becomes much more important for visibility in Search." For competitive commercial keywords, being in the top 5% of speed does not guarantee you rank — but being outside it more or less guarantees you don't.

Trust and perceived quality

This one is not in a spreadsheet, but it is the largest cost of all. A prospect landing on a premium hospitality website that takes 6 seconds to render its hero image has been shown, in the first 6 seconds of contact, that the brand does not care enough about them to invest in the basics of the experience. That prospect leaves before they see the price, the room, or the story. The brand never gets to make its case. This is the invisible cost — the leads you never knew you had.

The four causes of slow websites

In seven years and 21 rebuilds, I have found that ~95% of website slowness reduces to four causes. In descending order of frequency:

1. Media weight

Uncompressed or badly-formatted images and video. The typical premium-brand site ships hero images as 5MB JPEGs when they should be 80KB AVIFs. Video autoplays a 12MB MP4 when a 400KB WebM would look identical at that size on a phone screen. Media is the single largest opportunity on 90% of the sites I audit.

2. Third-party JavaScript

Chat widgets, heat maps, analytics stacks, marketing pixels, A/B testing frameworks, cookie banners. Each is "one small script." The composite is a JavaScript execution graph that blocks the main thread for 3+ seconds on a mid-tier Android phone. Third-party scripts are the second-largest contributor and the hardest one for non-technical owners to see, because the HTML looks clean.

3. CMS bloat

WordPress with 20+ plugins, Wix with 40+ apps installed, Shopify with dozens of theme customizations layered on top of each other. The CMS generates markup that no engineer would ever write by hand: 40kb of inline CSS per page, jQuery loaded three separate times, duplicated font declarations. The site owner sees a working page. The browser sees a 9,000-line HTML document.

4. Hosting and delivery

Shared PHP hosting in a single region, serving un-cached responses to every request. Your visitor in Sydney is waiting for a MySQL query to execute in a Frankfurt datacenter over a residential fiber connection. A modern static or edge-hosted site serves the same visitor from a Sydney edge node in under 40ms — the network cost drops to nearly zero.

How to measure your own site honestly

There are dozens of speed-testing tools. Three matter. Ignore the rest.

PageSpeed Insights (free, authoritative)

pagespeed.web.dev is Google's own tool, and Google is the source of truth for the ranking factor. Test mobile, not desktop. Test your most important pages — homepage, top landing page, top product/service page — not just the home. Take the median of three consecutive tests; single results can vary ±5 points from network noise.

Chrome DevTools Performance panel (free, diagnostic)

Load your site in Chrome, open DevTools (F12), select the Performance tab, and record a page load. The flame graph tells you exactly which scripts are blocking the main thread and for how long. This is where you find the third-party JavaScript problem.

WebPageTest (free tier, deep diagnostics)

webpagetest.org lets you test from real devices in real geographic locations on real network profiles (throttled 4G, throttled 3G). Use it when you want to see what your visitor in São Paulo on a mid-tier Android actually experiences, rather than what your MacBook on gigabit fiber shows you.

If you want me to run this measurement for you and send you a written+recorded audit within 24 hours, that is the free service I offer — with no expectation that you hire me afterward. If your site is already in the top 5%, I will tell you so and walk away.

Rebuild vs. optimize — the decision

Once you know your score, you face one decision: fix the existing site, or rebuild from scratch. Both are legitimate, both have their moment, and picking the wrong one is expensive.

When to optimize

  • Your PSI mobile score is between 70 and 90.
  • Your site is on a modern stack (Next, Astro, SvelteKit, Nuxt, headless Shopify) but was not disciplined about performance during build.
  • Your site's content and structure are working commercially — you just have a technical debt problem, not a strategy problem.
  • Your budget is under €2,500 and you can accept the ceiling of what optimization can deliver (~95 mobile is realistic; 100 is not).

When to rebuild

  • Your PSI mobile score is under 70. Below this threshold, optimization is bailing water out of a sinking boat.
  • You are on WordPress with 20+ plugins, or on any drag-and-drop builder (Wix, Squarespace, GoDaddy Builder). The platform is the ceiling.
  • The site was built more than 5 years ago and has been continuously accreted rather than periodically refactored.
  • Your content or brand strategy has changed and the site needs a structural rewrite anyway.
  • You want the guarantee of 95+. That number is not achievable on legacy stacks; it requires the rebuild path.

The honest truth about the middle ground

Optimization is often oversold as "the cheaper alternative to a rebuild." In my experience with 40+ audits, a site that scores under 60 will spend €4,000 in optimization consulting and end up at PSI 74. A rebuild would have taken that same site to 98 for €4,800. The maths only look favorable for optimization if you value the incumbent stack for reasons that have nothing to do with performance — usually a large custom back-end, a complex integration surface, or an editorial workflow that would take months to migrate.

What a top-5% rebuild actually looks like

The pattern I use for every rebuild is boring on purpose. Boring is fast. In 2026 the fastest websites in the world are architecturally simpler than the average WordPress install by two orders of magnitude.

Static or edge-rendered HTML

Every page is pre-built at deploy time (or rendered on-demand at the edge, close to the visitor). No PHP process, no database query, no application server. The visitor's browser requests a file, the CDN serves the file, the file renders. This is how the web was designed to work.

Minimal JavaScript, aggressive splitting

JavaScript is expensive. Every kilobyte of JS is parsed, compiled, and executed on the visitor's device — often a €200 Android phone with a slow CPU. A disciplined rebuild ships under 30KB of JS on the first page load. Interactivity that isn't visible on the first screen is loaded on demand. Frameworks like Astro (my default), Qwik, and modern SvelteKit are built around this principle.

Images in modern formats, correctly sized

Every image is served in AVIF or WebP, at the exact resolution the visitor's screen will render it, from a CDN. A hero image drops from 5MB to ~80KB with zero visible quality loss. This one change alone can move a site 30 PSI points.

No third-party scripts on the critical path

Analytics, chat, heat maps — all deferred until after the page is interactive. Cookie banners implemented as HTML+CSS, not as a JavaScript-loaded widget from an external vendor. Every third-party script is treated as a liability that must justify its performance cost.

Edge hosting from day one

My default stack is Cloudflare Pages + Workers. Your visitor in Tokyo is served from a Tokyo edge node in under 30ms. Your visitor in São Paulo from São Paulo. No single-region PHP host can compete with this geometrically; it is the difference between a truck and a plane.

How to hire someone (or how to check they know what they're doing)

If you are going to bring in a specialist rather than DIY, here is the diligence checklist. This is what I would ask any consultant or agency before signing anything.

  • Show me three sites you have shipped that score 95+ on PSI mobile today. Not "at launch." Today. If the score has drifted, the shop does not maintain a performance culture.
  • Written performance guarantee, or the money back. If there is no written performance clause, the shop is not confident. If the shop is not confident, you are the beta test.
  • Fixed public pricing. Performance is a commodity outcome (PSI is a public number). If the price is negotiable, the scope is negotiable, which means the outcome is negotiable.
  • Named person, not a team badge. Someone is doing the work. Ask who. Ask their portfolio. Ask what they will and will not do themselves.
  • Delivery timeline in weeks, not months. A performance rebuild that takes 6 months is a strategy rewrite dressed up as a technical project. Real rebuilds ship in 3 weeks.

For reference, my own answers to those five questions are: 21 sites currently at 95+ (see the proof grid); 100% money back if PSI 95+ is not achieved within 30 days of launch; €4,800 fixed, published, same for everyone; Michael Morakis, solo; 3 weeks from kickoff to launch.

Frequently asked questions

Is my website fast enough?

Open pagespeed.web.dev, paste your URL, run the mobile test. Any score under 90 has commercial consequences you are paying for whether you know it or not. Under 70 and you have a serious problem. Under 50 is a five-alarm fire.

What is a good PageSpeed score in 2026?

On mobile: 90+ is "good" per Google; 95+ is the top 5% of the web; 98+ is the top 1%. On desktop the bar is higher (most sites score 90+ on desktop for free), so mobile is the meaningful benchmark.

How long does it take to make a website fast?

A rebuild from a legacy platform to modern stack takes ~3 weeks in my workflow. Ad-hoc optimization of an existing modern stack can take 1–2 weeks. There is no honest 48-hour fix for a site that is genuinely slow architecturally.

Will making my website faster hurt SEO in the short term?

A well-executed rebuild preserves URL structure with 301 redirects for any URL that must change. Rankings often improve within 30–60 days as Core Web Vitals lift. Poorly executed rebuilds that break URLs can lose 20–40% of organic traffic. This is the single most important thing to get right, and it is why "cheap rebuild" is often the most expensive option in year one.

Do I need a rebuild if my site already scores 85?

Probably not. Sites in the 80–89 range usually have specific, isolable problems (a heavy chat widget, an unoptimized hero image, one bloated plugin) that a few days of targeted optimization can fix. Rebuilds are for the sub-70 range.

What does a fast website cost?

My published pricing for a rebuild starts at €4,800 for a standard marketing site up to ~15 pages, delivered in 3 weeks, with the 95+ PSI guarantee in writing. Larger platforms — e-commerce, multi-language, application-heavy — range from €8,000 to €12,000+. Ad-hoc optimization engagements start at €1,500. All pricing is public and non-negotiable.

Is speed the same as SEO?

Speed is one component of technical SEO, and it is the component with the highest ROI for most premium-brand sites because it is the one their competitors have all neglected. It is not a substitute for content, on-page optimization, or link authority — but it is the fastest lever to move, the most measurable, and the only one where the improvement lasts as long as the code does.

Can I do this myself?

If you are a developer with 2+ years of experience on Astro, Next, or SvelteKit, yes. The stack is well-documented and the community is generous. If you are not that person, hiring the specialization is faster and cheaper than becoming the specialization.

Ready to go from slow to top-5%?

I offer a free 24-hour audit of any site — PSI report, 5-minute screen-recorded walkthrough of what would need to change, and a written scope+quote if a rebuild makes sense. No obligation, no follow-up sequence, no newsletter unless you ask for one.

Request the free audit →

Or read the vertical-specific pages if you want to see the pattern applied to your industry: Insurance · Boutique fitness. More verticals shipping monthly.

Last updated 2026-08-12.

← Back to michaelmorakis.com